Assessing Diagnosis Approaches for Wireless Sensor Networks: Concepts and Analysis

Diagnosis is of great importance to wireless sensor networks due to the nature of error prone sensor nodes and unreliable wireless links. The state-of-the-art diagnostic tools focus on certain types of faults, and their performances are highly correlated with the networks they work with. The network administrators feel difficult on measuring the effectiveness of their diagnostic approaches and choosing appropriate tools so as to meet the reliability demand. In this work, we introduce the D-vector to characterize the property of a diagnosis approach. The D-vector has five dimensions, namely the Degree of Coupling, the Granularity, the Overhead, the Tool Reliability and the Network Reliability, quantifying and evaluating the effectiveness of current diagnostic tools in certain networks. We employ a skyline query algorithm to find out the most effective diagnosis approaches, i.e., skyline points (SPs), from five dimensions of all potential D-vectors. The selected skyline D-vector points can further guide the design of various diagnosis approaches. In our trace-driven simulations, we design and select tailored diagnostic tools for GreenOrbs, achieving high performance with relatively low overhead.